I have two questions:
1- As far as I know the default technique of MOSEK for CVX is interior-point method. How can I change that to the Simplex method?
2- In general, is it possible to have the feasible region of the decision variables besides the optimal value? Does MOSEK or any other solver provide that? I need the range of decision variables that they satisfy all of the constraints.
Thanks and All the best.
cvx_solver_settings( 'MSK_IPAR_OPTIMIZER', 'MSK_OPTIMIZER_FREE_SIMPLEX');
As far as I remember when you solve an LP even with interior point then CVX will give you the basic solution from the basis identification which follows the interior-point algorithm. But don’t take my word for it 100% and test yourself.
- No. It is not clear what you want. The feasible region is a polyhedron you already described when you specified the problem. If you want some ranges you can specify some variants of your problem that will compute them, such as “minimize/maximize x_1 subject to the constraints” or others along these lines.
Thanks for your reply.
- I tried that but I got error like:
Mosek error: MSK_RES_ERR_INV_OPTIMIZER (An invalid optimizer has been chosen for the problem.)
Optimal value (cvx_optval): NaN
I am using command “Lambad_max” in my code. I think the error is because of that. Because when I eliminate that line, the code woks again. What is the reason?
2- Yes I had used the technique you mentioned, but I was wondering if the solver can provide that efficiently.
All the best
lambda_max causes an SDP to be formulated. Mosek can’t solve that using Simplex.